Carbon emissions forecasting based on a new hybrid model under carbon reduction target: A case study of eastern region in China

A reasonable and accurate prediction technique constitutes the solid foundation for the achievement of carbon reduction goals and the formulation of relevant planning. Combined with the Radial Basis Function Neural Network and other machine learning techniques, a new hybrid technique in predicting CO₂ emitting trajectory, i.e. 3RGM model, is proposed. 3RGM can effectively conquer the defects of widely-used prediction methods (e.g. BPNN and STIRPAT model), such as easy falling into the local minimum value and low forecast precision. The assessment results indicate the prediction accuracy in carbon emission is improved by 25.88% at the largest and 8.10% on average through 3RGM, in comparison of the BPNN and STIRPAT models. Thereafter, through setting up three scenarios (BAU, HIGH, LOW), 3RGM is applied in forecasting 2020–2030 CO₂ emission of eastern region, the biggest CO₂ emitter in China. It is found that CO₂ emission in eastern region will peak in 2028 under three scenarios. GDP, energy consumption, and proportion of primary industry are the main driving forces for carbon emission. Additionally, Hebei and Shandong are the largest emitters, and all eastern provinces will achieve their carbon intensity goals, excluding Beijing under BAU, Zhejiang under BAU and LOW. Based on the prediction results, it is suggested that Zhejiang is adopting stimulus policies to energize social economy development, and Beijing is supposed to take stimulus or steady policies to alter development mode. In addition, curbing energy consumption and rationalizing industrial structure while stimulating economic growth, are important means to reduce carbon emissions in the eastern region. The 3RGM model enriches the CO₂ prediction methodology and gives technical support for government policy formulation from a forward-looking perspective.